As a teacher, the mechanics of learning have always fascinated me. They hit close to home. That’s why what follows —although grounded in years of teaching—comes out a bit more visceral than usual.
Consider this your disclaimer.
What I’m going to lay out is my view on technical learning (Web development) in the AI era: how most people use it poorly, why the dominant model is fundamentally flawed, and why generative AI might be the start of a new kind of online education that, until recently, sounded like science fiction.
A Single Model
Before looking ahead, let’s look at where we are.
Anyone who’s tried learning programming online ends up on some kind of video platform.
That’s the model. That’s the whole thing.
Video. Nothing more.
Maybe a bit of text. Maybe a couple of exercises. But video runs the show.
And I get it. I abused it for years too.
I’m not questioning the quality, there are brilliant examples despite the limitations of the medium.
The democratization of video transformed culture. It let knowledge travel to millions of people separated by oceans.
Truly revolutionary... in 2003.
But for Web development—or anything that requires actual practice—video hits a ceiling. It never really worked. Not fully.
From day one, it has been an inefficient model.
We simply didn’t have an alternative.
Until now.
A Single Proposal
Here’s the core problem: one method for everyone. A one-size-fits-all system.
The moment you step into the platform, your preferences and learning style are irrelevant. If you fit the mold, good for you. If you don’t… fuck you.
In a one-hour video (which might belong to a 50, 60, or 80-hour course), you’re just one more viewer in the crowd.
Zero personalization. Zero consideration for the actual learner.
- Same type of intelligence.
- Same starting point.
- Same personal circumstances.
- Same goals.
Can you learn this way? Yes—but despite the medium, not because of it. You can also have sex standing on your hands—possible, sure, but hardly the most effective method.
Content gets recycled. Metrics focus on quantity, not effectiveness.
Everyone who has created or consumed these courses knows something’s off. But the “solution” is always the same: more video, now in 4K with fancy editing.
Access to content stopped being the problem a long time ago.
The real challenge is consumption, dosage, assimilation, and—above everything—application.
The Only Tool
Does this mean video has no place in online learning?. No. It has a role. But not the whole damn stage.
When you only have one tool, you end up using it for everything—even when it’s not the best option.
Humans learn by watching others perform a task. Rizzolatti’s discovery of mirror neurons backs this up: watching someone do something activates the same brain regions as doing it yourself.
So yes, watching someone code, make decisions, debug in real time—that has pedagogical value.
But that’s only one phase. Real progress happens when you move from passive observation to deliberate practice and your own mistakes.
First you absorb patterns. Then you face problems alone, apply what you saw, and fail your way through.
Video is valid—but as one tool among many, not the only one.
The Science Is Clear
We’ve known for decades how humans learn and what conditions make learning stick. None of this is new.
And yet the “ed-tech” response has always been the same:
More video.
While in-person environments use some of these principles (personalization, active recall, etc.), virtual classrooms still teach in a way that largely ignores what actually benefits the learner.
Scientific Principles of Learning
- We forget fast unless we actively practice. The Ebbinghaus forgetting curve (1885). If you don’t retrieve information, it evaporates. The fix is proven: spaced repetition. Revisit what you learn at increasing intervals—with active challenges, not passive review.
- It’s not what you read—it’s what you retrieve. The Testing Effect (Karpicke & Roediger, 2008). Memory strengthens through recall, not re-reading. That’s why well-designed quizzes and microtests are essential.
- Immediate feedback matters more than you think. Not just “right or wrong.” Fast, concrete feedback (Hattie & Timperley, 2007) accelerates correction. Waiting a week for feedback is archaic. Learning happens in the moment.
- Deliberate practice is not doing more—it’s repeating what you don’t know. Ericsson (1993) showed this clearly: experts don’t repeat the easy stuff. They push where they’re weak. Without a system that forces you into discomfort, progress stalls.
- Learning builds on what you already know. Vygotsky, Novak, and cognitive psychology: new knowledge only sticks when it connects to prior knowledge. Isolated learning is an illusion.
- Personalization multiplies outcomes. Bloom’s 2 Sigma Problem (1984): personalized feedback places students two standard deviations above those receiving generic instruction.
- There’s no motivation without autonomy. Deci & Ryan (1985): badges and points have limited impact. What really matters is autonomy, discipline, and control over your learning.
- Community accelerates progress. Bandura, Lave & Wenger: social learning multiplies results. Explaining, discussing, and observing others’ mistakes improves your own understanding far faster than working solo.
Now you know why you forget what you study, why you have abandoned courses collecting dust, or why—even if you finish them you still feel like nothing really “landed.”
Inefficient Use of AI Assistance
Generative AI could be a key factor—maybe the factor—in online learning. Used well, it can reshape distance education. I speak about programming, but this applies anywhere knowledge is the core currency.
Think about why video became the default format. Because for decades, there was no viable alternative.
What’s the point in philosophizing about advanced tools when all you have is a stick, a rope, and a rock?
But now the alternative exists.
Science gave us the blueprint—personalization, rapid feedback, deliberate practice, level-adjustment, active recall—and today we can actually implement, measure, refine, and scale it.
What used to be sci-fi is now reachable.
The problem? Most people still use this tech like amateurs. They copy-paste into ChatGPT like they did into StackOverflow. No process. No intention. No system.
The potential exists, but potential alone means nothing.
AI won’t magically teach you. The learner needs intention, and the educational system needs structure.
A machine-human loop:
- The human sets the path, curates content, establishes criteria.
- The machine amplifies, personalizes, adapts feedback, and accelerates practice cycles.
Without a system, AI is just another hype-driven shortcut.
With a system, it becomes a real catalyst for transformation.
At least, the possibility is there.
Do You Remember What Learning Is?
What do you think when you see those gym posters with people lifting weights or cycling with a big smile?
Fake. Staged. Unreal.
Anyone who has actually trained knows that’s not how it works. They’re actors. It’s a façade. Nothing about it is real.
No substance, just appearance.
Online learning has followed the same pattern. “Master X,” “Learn Y,” “Become an expert in Z.”
Again, reality refuses to align with the marketing. The process isn’t clean or painless. You don’t breeze through hours of content with that YouTuber stupid grin.
Maybe lesson one feels good. By lesson three or four, doubts pile up, concepts blur, and details slip through the cracks.
Real learning gets stuck. It clumps. It resists.
It’s push and pull—and that’s exactly how it should be.
It’s not neat. It’s not always fun. It’s messy and stressful, but predictable because it follows well-understood rules.
And it works because humans are built for it.
We’re wired to do, not just to watch.
Keep that in mind the next time Udemy goes on sale.